Re: [linux-audio-dev] Juce now has ALSA support!

On Friday 03 March 2006 02:38, Lee Revell wrote: Well, there is a reason why all "serious" audio APIs use the callback model. It's the only model that really works for low latency audio. If you want to do serious real time audio, you'll have to get your head around this model anyway. (You need to keep the CPU load steady, and generating N samples every cycle rather than larger blocks "every now and then" is the first step.) If you're not doing real time stuff, you can wrap any API with whatever type of API you like, as the buffering that is sometimes required, won't matter. Indeed, there are some "latency reduction" tricks you can play with an mmap() interface, but those are inefficient, messy hacks that you are (or rather, were) forced to use to get usable latency in games on some platforms. Not really usable for musical applications, and the method doesn't fit well into engines with master effects and the like. Callbacks on a proper OS are simpler, more reliable and more efficient. A read()/write() interface is basically just a (more or less) buffered wrapper over a real interface. I frankly don't understand why one would want to use something like it for anything remotely resembling real time audio I/O. All it does is confuse matters WRT how much buffering you actually have, especially when you're doing full duplex audio. It may be handy in non real time applications (sometimes it's easier to let the algorithm control the block sizes), but as those aren't sensitive to additional buffering, you can just wrap the real API with whatever you like. No reason to break or complicate the real API to support that. //David Olofson - Programmer, Composer, Open Source Advocate .------- http://olofson.net - Games, SDL examples -------. | http://zeespace.net - 2.5D rendering engine | | http://audiality.org - Music/audio engine | | http://eel.olofson.net - Real time scripting | '-- http://www.reologica.se - Rheology instrumentation --'